testing-llm（测试 LLM）

LLM & AI Testing Patterns

Patterns and tools for testing LLM integrations, evaluating AI output quality, mocking responses for deterministic CI, and applying agentic test workflows (planner, generator, healer).

Quick Reference

When to Use This Skill

• Testing code that calls LLM APIs (OpenAI, Anthropic, etc.)
• Validating RAG pipeline output quality
• Setting up deterministic LLM tests in CI
• Building evaluation pipelines with quality gates
• Applying agentic test patterns (plan -> generate -> heal)

LLM Mock Quick Start

Mock LLM responses for fast, deterministic unit tests:

from unittest.mock import AsyncMock, patch
import pytest

@pytest.fixture
def mock_llm():
    mock = AsyncMock()
    mock.return_value = {"content": "Mocked response", "confidence": 0.85}
    return mock

@pytest.mark.asyncio
async def test_with_mocked_llm(mock_llm):
    with patch("app.core.model_factory.get_model", return_value=mock_llm):
        result = await synthesize_findings(sample_findings)
    assert result["summary"] is not None

Key rule: NEVER call live LLM APIs in CI. Use mocks for unit tests, VCR.py for integration tests.

DeepEval Quality Quick Start

Validate LLM output quality with multi-dimensional metrics:

from deepeval import assert_test
from deepeval.test_case import LLMTestCase
from deepeval.metrics import AnswerRelevancyMetric, FaithfulnessMetric

test_case = LLMTestCase(
    input="What is the capital of France?",
    actual_output="The capital of France is Paris.",
    retrieval_context=["Paris is the capital of France."],
)

assert_test(test_case, [
    AnswerRelevancyMetric(threshold=0.7),
    FaithfulnessMetric(threshold=0.8),
])

2026 library updates (DeepEval 2.3, RAGAS 1.2)

DeepEval 2.3 introduces self-explaining scores — every metric now emits a reason field alongside the numeric score, so a failing CI build gets a human-readable explanation without a second LLM call:

metric = AnswerRelevancyMetric(threshold=0.7, include_reason=True)
metric.measure(test_case)
print(metric.score, metric.reason)
# 0.62  "Response addresses the topic but omits the date asked for."

RAGAS 1.2 ships dynamic recalibration — when the grader model drifts (e.g. GPT-5.2 → future Gemini 3.1), RAGAS records the shift and adjusts the threshold so historical scores stay comparable across evals:

from ragas.evaluation import evaluate
from ragas.metrics import faithfulness, context_recall

result = evaluate(
    dataset,
    metrics=[faithfulness, context_recall],
    recalibrate=True,   # 1.2+ — normalizes against the grader baseline
)

Bump floors: deepeval >= 2.3, ragas >= 1.2. Older releases silently drop the reason/recalibrate kwargs.

Quality Metrics Thresholds

Structured Output Validation

Always validate LLM output with Pydantic schemas:

from pydantic import BaseModel, Field

class LLMResponse(BaseModel):
    answer: str = Field(min_length=1)
    confidence: float = Field(ge=0.0, le=1.0)
    sources: list[str] = Field(default_factory=list)

async def test_structured_output():
    result = await get_llm_response("test query")
    parsed = LLMResponse.model_validate(result)
    assert 0 <= parsed.confidence <= 1.0

VCR.py for Integration Tests

Record and replay LLM API calls for deterministic integration tests:

@pytest.fixture(scope="module")
def vcr_config():
    import os
    return {
        "record_mode": "none" if os.environ.get("CI") else "new_episodes",
        "filter_headers": ["authorization", "x-api-key"],
    }

@pytest.mark.vcr()
async def test_llm_integration():
    response = await llm_client.complete("Say hello")
    assert "hello" in response.content.lower()

Agentic Test Workflow

The three-agent pattern for end-to-end test automation:

Planner -> specs/*.md -> Generator -> tests/*.spec.ts -> Healer (auto-fix)

1. Planner (references/planner-agent.md): Explores your app, produces Markdown test plans from PRDs or natural language requests. Requires seed.spec.ts for app context.
2. Generator (references/generator-agent.md): Converts Markdown specs into Playwright tests. Actively validates selectors against the running app. Uses semantic locators (getByRole, getByLabel, getByText).
3. Healer (references/healer-agent.md): Automatically fixes failing tests by replaying failures, inspecting the DOM, and patching locators/waits. Max 3 healing attempts per test.

Edge Cases to Always Test

For every LLM integration, cover these paths:

• Empty/null inputs -- empty strings, None values
• Long inputs -- truncation behavior near token limits
• Timeouts -- fail-open vs fail-closed behavior
• Schema violations -- invalid structured output
• Prompt injection -- adversarial input resistance
• Unicode -- non-ASCII characters in prompts and responses

See checklists/llm-test-checklist.md for the complete checklist.

Anti-Patterns

Related Skills

• ork:testing-unit — Unit testing fundamentals, AAA pattern
• ork:testing-integration — Integration testing for AI pipelines
• ork:golden-dataset — Evaluation dataset management